Neuronal morphology in the African elephant (Loxodonta africana) neocortex

Page created by Diana Gomez
 
CONTINUE READING
Neuronal morphology in the African elephant (Loxodonta africana) neocortex
Brain Struct Funct
DOI 10.1007/s00429-010-0288-3

 ORIGINAL ARTICLE

Neuronal morphology in the African elephant
(Loxodonta africana) neocortex
Bob Jacobs • Jessica Lubs • Markus Hannan •
Kaeley Anderson • Camilla Butti • Chet C. Sherwood           •

Patrick R. Hof • Paul R. Manger

Received: 14 August 2010 / Accepted: 15 October 2010
! Springer-Verlag 2010

Abstract Virtually nothing is known about the mor-               supragranular pyramidal neurons, elephant superficial
phology of cortical neurons in the elephant. To this end, the    pyramidal neurons exhibited similar overall basilar den-
current study provides the first documentation of neuronal       dritic length, but the dendritic segments tended to be longer
morphology in frontal and occipital regions of the African       in the elephant with less intricate branching. Finally, ele-
elephant (Loxodonta africana). Cortical tissue from the          phant aspiny interneurons appeared to be morphologically
perfusion-fixed brains of two free-ranging African ele-          consistent with other eutherian mammals. The current
phants was stained with a modified Golgi technique.              results thus elaborate on the evolutionary roots of Afro-
Neurons of different types (N = 75), with a focus on             therian brain organization and highlight unique aspects of
superficial (i.e., layers II–III) pyramidal neurons, were        neural architecture in elephants.
quantified on a computer-assisted microscopy system using
Neurolucida software. Qualitatively, elephant neocortex          Keywords Dendrite ! Morphometry ! Dendritic spine !
exhibited large, complex spiny neurons, many of which            Golgi method ! Brain evolution
differed in morphology/orientation from typical primate
and rodent pyramidal neurons. Elephant cortex exhibited a
V-shaped arrangement of bifurcating apical dendritic             Introduction
bundles. Quantitatively, the dendrites of superficial pyra-
midal neurons in elephant frontal cortex were more com-          In recent years, elephants have received considerable
plex than in occipital cortex. In comparison to human            attention because of their complex social structure and
                                                                 cognitive abilities (Poole and Moss 2008; Hart and Hart, in
                                                                 press). Nevertheless, despite elephants’ status as the largest
B. Jacobs (&) ! J. Lubs ! M. Hannan ! K. Anderson
                                                                 living terrestrial mammal, very little is known about the
Laboratory of Quantitative Neuromorphology, Psychology,
The Colorado College, 14 E. Cache La Poudre,                     brain of this species. Prior to 2001, only a few original
Colorado Springs, CO 80903, USA                                  articles had specifically focused on the elephant central
e-mail: BJacobs@ColoradoCollege.edu                              nervous system (Cozzi et al. 2001). Although the gross
                                                                 anatomy of the *5 kg adult elephant brain has been
C. Butti ! P. R. Hof
Department of Neuroscience and Friedman Brain Institute,         investigated in the last decade (Kupsky et al. 2001; Hakeem
Mount Sinai School of Medicine, One Gustave L. Levy Place,       et al. 2005, 2009; Shoshani et al. 2006; Manger et al. 2010),
New York, NY 10029, USA                                          virtually nothing is known about the microstructural mor-
                                                                 phology of its neurons. Indeed, until recently, the only
C. C. Sherwood
Department of Anthropology, The George Washington                apparent insight into elephant cortical neuromorphology
University, 2110 G Street, NW, Washington, DC 20052, USA         was a single camera lucida drawing from an Indian elephant
                                                                 (Elephas maximus; Barasa and Shochatovitz 1961). Rea-
P. R. Manger
                                                                 sons for this poverty of data are many (Cozzi et al. 2001),
Faculty of Health Sciences, School of Anatomical Sciences,
University of the Witwatersrand, 7 York Road, Parktown,          but include a propensity for neuroscientists to focus on the
Johannesburg 2193, South Africa                                  rodent and primate species commonly used in biomedical

                                                                                                                    123
Brain Struct Funct

research (Manger et al. 2008), and a lack of well-preserved            Afrotheria contains many neural plesiomorphies in com-
elephant brain tissue suitable for histological analysis. This         parison with other placental mammal species (Sherwood
latter issue was recently addressed by Manger et al. (2009),           et al. 2009). Thus, examination of neuronal morphology in
who were able to fix the brains of wild African elephants              elephants is essential for understanding broader phyloge-
(Loxodonta africana) by postmortem, carotid cannulation-               netic patterns of neuromorphological evolution in Afrothe-
perfusion. The quality of this tissue facilitated the present          rians and provides data concerning the scaling of neuronal
investigation, which qualitatively and quantitatively                  somatodendritic geometry with increasing body and brain
explores, for the first time, the neuromorphology of frontal           size (Tower 1954; Haug 1987; Hart et al. 2008).
and occipital cortices in the African elephant.                            Both cognitive and sociobehavioral investigations of
   Elephants belong to the order Proboscidea, which                    elephants reinforce the view that such a large brain is
emerged approximately 60 million years ago from within                 associated with extensive abilities (Roth and Dicke 2005;
the eutherian superorder Afrotheria (Sukumar 2003;                     Hart and Hart, in press). The cognitive repertoire of ele-
Shoshani and Tassy 2005; Gheerbrant and Tassy 2009;                    phants includes elementary tool construction/usage (Hart
Fig. 1). Extant afrotherian species (e.g., elephants, mana-            et al. 2001), impressive spatial and temporal memory
tees, dugongs, hyraxes, aardvarks, golden moles, tenrecs,              (Markowitz et al. 1975; Hart et al. 2008), creative problem
and elephant shrews) exhibit significant diversity in terms of         solving skills (Bates et al. 2008b), the ability to classify
body and brain mass, neural organization, and life history             human ethnic groups by odor and garment color (Bates
(Gravett et al. 2009; Pieters et al. 2010). Because the afro-          et al. 2007a), and the ability to locate out-of-site family
therian clade represents a major adaptive radiation within             members by smell (Bates et al. 2007b). Social interactions
Eutheria, it is in a position to provide insight into mamma-           in elephants are underscored by several documented abil-
lian brain evolution, with current evidence suggesting that            ities: ultra-low frequency sound communication over both
                                                                       local and long distances (Poole 1999; Garstang 2004; Soltis
                                                                       2009), vocal learning (Poole et al. 2005), long-term care of
                                                                       their young (Bates et al. 2008b), individual identification of
                                                                       conspecifics (Bates et al. 2007b), and targeted helping
                                                                       behavior (Hart et al. 2008). These abilities suggest ele-
                                                                       phants may be capable of empathy (Bates et al. 2008a, b)
                                                                       and self-recognition, including theory of mind (Plotnik
                                                                       et al. 2006), all of which crucially contribute to their
                                                                       complex social networks (McComb et al. 2000, 2001).
                                                                           Unfortunately, besides the recent documentation of von
                                                                       Economo neurons in the anterior insular cortex of elephants
                                                                       (Hakeem et al. 2009), one can only extrapolate the neural
                                                                       foundations for these cognitively demanding abilities. For
                                                                       example, based on the low density of neurons in their
                                                                       cerebral cortex, it has been suggested that the elephants’
                                                                       cognitive faculties may be concentrated in long-term pro-
                                                                       cessing (Hart and Hart 2007; Hart et al. 2008), although fine
                                                                       sensorimotor integration is certainly required to guide the
                                                                       trunk (Hoffmann et al. 2004). Apart from such speculation,
                                                                       there are currently no data that allow for direct examination
                                                                       of the computational biophysics of elephant cortical neu-
                                                                       rons. However, in the prototherian echidna, whose ancestry
                                                                       is close to the root of the mammalian phylogenetic tree
                                                                       (Fig. 1), a large proportion of pyramidal neurons contain so-
                                                                       called ‘‘atypical’’ features, such as bifurcated apical den-
Fig. 1 Taxonomy of the elephant (Loxodonta africana) and a
sampling of related species mentioned in the present paper. During
                                                                       drites, inverted somata, poorly developed basilar skirts, and
the late Cenozoic era, Eutheria (placental mammals) diverged into      a lack of terminal bouquets in layer I (Hassiotis et al. 2003,
two phylogenetic groups: Atlantogenata and Boreoeutheria (Wildman      2005). This suggests that there may be significant phyloge-
et al. 2007). Atlantogenata subsequently diverged into two sister      netic diversity in the morphology of spiny projection neu-
clades, Xenarthra (e.g., sloths, anteaters, and armadillos) and
Afrotheria, whereas Boroeutheria diverged into Laurasiatheria (e.g.,
                                                                       rons of the cerebral cortex. As such, based on its afrotherian
cetaceans, carnivores, and ungulates) and Euarchontoglires (e.g.,      origin, we expect a variety of spiny neuron morphologies in
primates and rodents)                                                  the elephant that differs from the euarchontoglire branch of

123
Brain Struct Funct

Fig. 2 a Photograph of elephant LA1; b superior and c inferior views c
of the elephant brain illustrating the relative position of sampled
tissue blocks (shaded areas) for the occipital and frontal regions,
respectively. Note that, in c, the right hemisphere’s olfactory bulb has
been removed in order to view the orbital gyri and facilitate removal
of the tissue block. Scale bar 5 cm

placental mammals, mostly typified by anthropoid primate
and murid rodent models, and thus appear ‘‘atypical’’. In
contrast, the morphological classes for aspiny interneurons
are more uniform across eutherian mammals (Hof et al.
1996, 1999; DeFelipe et al. 2002; Hassiotis and Ashwell
2003; Hof and Sherwood 2005). Thus, we expect aspiny
interneurons in elephants to resemble more closely the types
that are common in the mammalian cerebral cortex.
   By extension, neuromorphological diversity in the ele-
phant implies a diverse cytoarchitecture, which would
contradict the long-held, but problematic claim that large
brains—such as in cetaceans—lack histological complexity
(Kesarev 1975; Glezer et al. 1988). In fact, recent studies on
cetacean cerebral cortices have revealed a complex cytoar-
chitecture, comparable to that observed in many mammals,
including primates (Hof et al. 2005; Hof and Van der Gucht
2007; Butti et al. 2009; Butti and Hof 2010). The present
study examines such neuromorphological diversity by
exploring regional variation in neuron morphologies across
all layers of the elephant cortex and by examining quanti-
tative differences in superficial (i.e., layers II–III) pyramidal
neurons between frontal and occipital cortical regions. As is
the case with human supragranular pyramidal neurons, these
superficial pyramidal neurons also tended to stain clearly
with the rapid Golgi technique (Jacobs et al. 2001). Both
functional (Foxe and Simpson 2002) and morphological
data in primates (Elston and Rosa 1997; Jacobs et al. 2001)
suggest regional variation in cortical processing, generally
with successively more complex processing as one moves
rostrally. Greater processing demands on neurons appear to
be reflected in the increasing complexity of their dendritic
arrays (Elston 2003, 2007; Jacobs and Scheibel 2002).
Although it is not possible to know which functional cortical
areas are contained in the frontal and occipital samples in the
present study, examination of dendritic complexity, as
revealed by superficial pyramidal neurons, may indicate
regional variation of integrative functions in the elephant.

Materials and methods

Specimens

Two (LA1 and LA3; Fig. 2a), free ranging, solitary male
African elephants (Loxodonta Africana; approximate age

                                                                           123
Brain Struct Funct

20–30 years) scheduled for population management culling       focus on superficial (i.e., layers II–III) pyramidal neurons
were humanely killed as described in Manger et al. (2009).     (mean soma depth = 794 ± 131 lm) in the two cortical
In situ perfusion-fixation of the brains was conducted by      regions. Because the Golgi impregnation was not uniform,
removal of the head, flushing of the cranium with cold         more neurons appropriate for quantification were sampled
saline, and intra-carotid perfusion with 4% paraformalde-      from LA1 (n = 46) than from LA3 (n = 29). A total of 40
hyde in 0.1 M phosphate buffer, resulting in relatively        superficial pyramidal neurons were traced in the occipital
short autolysis times (LA1: 160 min; LA3: 110 min). The        cortex (15 from LA1; 5 from LA3) and in the frontal cortex
brains were then removed from the skull, placed in the         (20 from LA1); of these, 16 had relatively complete apical
same cold fixative, weighed (LA1: 5,145 g, LA3: 4,835 g),      dendrites, which were analyzed only descriptively.
and stored in 4% paraformaldehyde in 0.1 M phosphate           Although pooling data across animals was not an optimal
buffer for 72 h. Small tissue blocks containing the cortical   solution, an ANOVA comparing occipital total dendritic
regions of interest were removed and stored in 0.1%            length in superficial neurons between LA3 and LA1 found
sodium azide in 0.1 M phosphate buffer saline at 4"C for       no significant difference, suggesting this as an acceptable
8 months before staining. The present research protocol        compromise for subsequent analyses. A broader sampling
was approved by the Colorado College Institutional             of neuron morphology resulted in 16 additional, deep layer
Review Board (#H94-004) and by the University of the           III to layer VI pyramidal neurons being traced in occipital
Witwatersrand Animal Ethics Committee (Manger et al.           cortex (10 from LA1; 6 from LA3), and 19 in frontal cortex
2009).                                                         (1 from LA1; 18 from LA3).
                                                                  Previously established selection criteria for human su-
Tissue selection                                               pragranular pyramidal neurons (Jacobs et al. 1997, 2001;
                                                               Anderson et al. 2009) were adapted in the present study.
Frontal and occipital tissue blocks (1–2 cm, perpendicular     Briefly, selected neurons were required to be relatively
to the long axis of the sampled gyri) were removed from        isolated and unobscured, to appear fully impregnated with
the right hemisphere. Because no cytoarchitectural criteria    a soma roughly centered within the 120 lm-thick section,
exist in the elephant for identification of these cortical     and to have as complete dendritic arbors as possible. In
regions, localization was based solely on gross examination    order to create a relatively homogeneous cell population
of elephant brains (Haug 1970; Hakeem et al. 2005;             for superficial pyramidal neurons, a running average of
Shoshani et al. 2006) and on reports from other species        soma depth from the pial surface was maintained and cells
(e.g., giraffes: Badlangana et al. 2007; cetaceans: Hof and    were chosen to preserve similar averages across both cor-
Van der Gucht 2007). Occipital tissue was removed from         tical areas.
the parieto-occipital contour close to the midline, approx-       Cells were quantified along x-, y-, and z-coordinates
imately 3–5 cm immediately anterior to the cerebellum,         using an Olympus BH-2 microscope under a Planachromat
although it should be noted that the elephant occipital lobe   409 (NA = 0.70) dry objective interfaced with a Neu-
is not as clearly demarcated as in humans (Shoshani et al.     rolucida software system (MBF Bioscience, Williston, VT,
2006; Fig. 2b). Frontal tissue was removed from the            USA). The system utilized a MicroFire Digital CCD 2-
anterior portion of what appeared to be the orbital gyri,      Megapixel camera (Optronics, Goleta, CA, USA), which
3–5 cm from the midline, in a position superior to the         was mounted on a trinocular head. The camera image was
anterior third of the olfactory bulb (Fig. 2c).                viewed on a Dell E248WFP 24-inch LCD monitor with
   Tissue blocks were trimmed to 3–5 mm in thickness,          1,920 9 1,200 resolution. First, the soma was traced at its
coded to prevent experimenter bias, and processed by a         widest point in the two-dimensional plane in order to
modified rapid Golgi technique (Scheibel and Scheibel          provide an estimate of its cross-sectional area. Dendrites
1978b). To be consistent with previous studies (Jacobs         were then traced somatofugally in their entirety, account-
et al. 1997, 2001), processed tissue was serially sectioned    ing for dendritic diameter and quantifying all visible
at 120 lm with a vibratome. Tissue blocks adjacent to          spines, regardless of spine type or differences in spine
those removed for Golgi analysis were examined with a          length. Dendritic processes were not followed into adjacent
routine Nissl stain to establish laminar and cortical depth.   sections; as such, broken tips and unclear terminations
Thirty depth measurements for each layer were averaged         were identified as incomplete endings.
across 12 sections from each cortical region.                     Neurons were traced by three observers (BJ, JL, and
                                                               KA). Intrarater reliability was assessed by having each
Neuron selection and quantification                            rater trace the same dendritic branch, with somata and
                                                               spines, 10 times. Tracings varied little, as indicated by
Analyses qualitatively and quantitatively characterized        the relatively low coefficient of variation for soma size
elephant neocortical neurons (N = 75), with a particular       (2.8%), total dendritic length (TDL 2.8%), and dendritic

123
Brain Struct Funct

spine number (DSN 3.9%). Intrarater reliability was also
tested with a split plot design (a = 0.05), which revealed
no significant difference within tracers between the first
five tracings and the last five tracings of a dendritic
system. For interrater reliability, tracers were normed by
comparing their tracings of 10 dendritic systems to the
same tracings completed by the primary investigator (BJ).
Pearson’s product correlations across soma size, TDL,
and DSN (defined below) averaged 0.97, 0.96, and
0.96, respectively. An analysis of variance (ANOVA;
a = 0.05) indicated that there was no significant differ-
ence among tracers in the three measures. Finally, the
primary investigator reexamined all final tracings to
ensure accuracy.

Cell descriptions and dependent dendritic/spine
measures                                                      Fig. 3 Photomicrographs of Nissl-stained cortex from frontal (a) and
                                                              occipital (b) cortex with labeled layers. Note that the elephant appears
                                                              to lack a well-developed layer IV. Scale bar 1 mm
Deep pyramidal neurons were qualitatively described
according to nomenclature found in previous neuro-
                                                              Table 1 Laminar and cortical depth from the pial surface in elephant
morphological research (e.g., Ngowyang 1932; Ferrer et al.    cortex (lm)
1986a, b). Descriptions for each neuron used qualitative
                                                                                                Frontal region       Occipital region
criteria, such as cortical location, size, dendritic field
patterns, presence of spines, and soma characteristics.       Layer I/II junction                 560 ± 14             519 ± 20
Nevertheless, because axons were not visible and because      Layer II/III junction               730 ± 13             675 ± 19
intermediates between different neuronal types exist, these   Layer III/V junction              1,495 ± 14           1,643 ± 38
designations remain tentative.                                Layer V/VI junction               2,034 ± 21           2,098 ± 42
   For quantitative measures, a centrifugal nomenclature      Gray/white matter junction        2,828 ± 41           2,737 ± 55
was used (Bok 1959; Uylings et al. 1986): dendritic
                                                              Values represent mean ± SEM
branches arising from the soma are first-order segments
until they bifurcate into second-order segments, which
branch into third-order segments, and so on. In addition to   Independent variables and statistical analyses
noting soma size (as apparent surface area in lm2) and
soma depth from the pial surface, five previously estab-      For superficial pyramidal neuron comparisons, only basilar
lished measures (Jacobs et al. 2001) were used to analyze     dendrites were examined inferentially because apical den-
each neuron. Total dendritic length (TDL, lm) refers to       drites were often incomplete. Basilar dendritic data were
the summed length of all dendritic segments. Mean seg-        aggregated by neuron (CELL), and brain region (AREA).
ment length (MSL, lm) is the average length of each           The resulting dataset was analyzed using a one-way anal-
dendritic segment. Dendritic segment count (DSC) rep-         ysis of variance (ANOVA; a = 0.05) to investigate the
resents the number of dendritic segments. Dendritic spine     effects of area (i.e., frontal and occipital) on each depen-
number (DSN) refers to the total number of spines on          dent measure for the basilar skirt.
dendritic segments. Dendritic spine density (DSD) is the
average number of spines per micron of dendritic length.
Additionally, dendritic diameter throughout the cell was      Results
determined, thus providing a sixth measure: dendritic
volume (Vol, lm3). Although tested independently, many        Overview
of these measures are inherently interrelated. Finally,
dendritic branching patterns of all traced neurons were       Nissl-stained sections indicated that the frontal cortex was
examined using Sholl analysis (Sholl 1953), which             slightly thicker than the occipital cortex, with both regions
quantified the number of dendritic intersections at 20-lm     exhibiting a relatively thick layer III and lacking a distinct
intervals radiating centrifugally from the soma. This         layer IV (Fig. 3; Table 1). The majority of cells traced
analysis assisted in the characterization of apical and       resided in layers III and V. In Golgi preparations, the
basilar dendrites.                                            dominant characteristic across both occipital and frontal

                                                                                                                          123
Brain Struct Funct

Fig. 4 Low magnification photomicrograph of elephant frontal              magnopyramidal-taproot or matriarch neuron (D, also seen in
cortex illustrating a wide diversity of neuronal morphologies: inverted   Fig. 5J); and fork neuron (F, also seen in Fig. 5F). Pial surface is
pyramidal neurons (A, C); superficial pyramidal neurons (B, E);           at the top of the photomicrograph. Scale bar 500 lm

regions was the vast heterogeneity of spiny cell types                    of approximately 1,500 lm in layer III of the occipital
(Fig. 4). Generally, superficial layers contained many                    lobe, projected an ascending, typically bifurcating, apical
small, bifurcating variations of the pyramidal cell with                  process, often exceeding 1,200 lm in length. Their somata
thick apical dendrites that ramified as they ascended toward              varied in size (Table 2) and shape, from small and trian-
the pial surface. Deeper neurons possessed broad basilar                  gular to very large and fusiform. The basilar skirt extended
skirts and often extended multiple widely diverging apical                in all directions to produce a circular receptive area. The
dendrites. In some cells, a prominent basilar dendrite or                 magnopyramidal cells were generally less spiny than their
taproot penetrated the white matter. Aspiny neurons tended                superficial counterparts, with a DSD between 0.33 and 0.47
to possess long dendrites that bifurcated near the soma or                (Table 2). Sholl analysis revealed a high density of basilar
not at all, and were located in upper layer III or lower layer            dendrites similar in size and distribution to those of
V. Tracings of individual cell types are provided in Figs. 5,             superficial pyramidal cells, although the apical dendrite
6, and 7 with quantitative data provided in Table 2. It                   was always longer (Fig. 8a).
should be noted that laminar boundaries are not indicated                    An apparent variation of the above was the magnopy-
in Figs. 5, 6, and 7 because, given the considerable vari-                ramidal-taproot neurons (n = 4, Figs. 4D, 5I–L, 9), which
ability across individual sections, neurons of similar depth              were located in layers III and V of the frontal lobe. Because
may not necessarily be located in the same cortical layer                 of their prominent appearance in Golgi-stained sections
across different sections. These descriptions, based on                   and their trunk-like taproot, we named these ‘‘matriarch’’
examination of all Golgi-stained sections as well as the                  neurons (referring to the matriarch’s prominent role in
quantified neurons themselves, are presented below with                   elephant society). They exhibited extensive dendritic
superficial pyramidal neurons being considered separately                 branching, which suggests a broad sampling of cortical
because they were more extensively analyzed.                              information (Brown et al. 2008). The matriarch neuron’s
                                                                          fusiform soma extended an apical process for up to 900 lm
Spiny neurons                                                             that often bifurcated during its ascent to the pial surface. A
                                                                          basilar skirt projected densely and uniformly in all direc-
Deeper pyramidal neurons with ascending apical dendrites                  tions. Unique to this neuron was the addition of a complex,
exhibited several variations: magnopyramidal (with and                    descending taproot that tended to bifurcate widely into
without a taproot), multiapical, and fork neurons. Magno-                 smaller segments while the main branch extended for long
pyramidal neurons (n = 3, Fig. 6F–H), located at a depth                  distances (up to 1 mm) toward the underlying white matter.

123
Brain Struct Funct

Fig. 5 Neurolucida tracings of spiny neurons in the elephant frontal    magnopyramidal-taproot or matriarch neurons (I–L); multiapical
cortex presented to indicate their relative soma depths from the pial   pyramidal neuron (M). The quantitative dependent measures for all
surface (in lm): crab-like neuron (A); superficial pyramidal neurons    neurons are provided in Table 2. Scale bar 100 lm
(B–E); fork neuron (F); inverted pyramidal neuron (G, H);

DSD ranged from 0.32 to 0.54 (Table 2). In the Sholl                       A fork neuron (or Gabelzelle; n = 1, Figs. 4F, 5F),
analysis, the matriarch neuron (Fig. 8c) was second only to             originally described by Ngowyang (1932) in the human
the neurogliaform cell (Fig. 7L) in terms of overall den-               frontoinsular cortex, was found at the boundary between
dritic density.                                                         layers III and V of the frontal cortex. It possessed a split
   Multiapical pyramidal neurons (n = 3, Figs. 5M, 6K–L),               soma that tapered into two thick, ascending apical den-
present both in frontal and in occipital cortex, were among             drites 600 lm in length. Basilar projections extended
the deepest of the spiny cells, located predominantly in                obliquely as fine, parallel dendritic processes. DSD was
layer V at a depth of 1,500–1,700 lm (Table 2). Although                relatively high at 0.58 (Table 2). Its axon emerged proxi-
often incomplete due to sectioning, multiple, prominent                 mally from a basilar dendrite and appeared to descend. The
apical dendrites extended from a spherical cell body to                 fork neuron was unusual in that, like the horizontal pyram-
ascend symmetrically toward the pial surface. Basilar                   idal cell, its apical and basilar branches were similar in
dendrites extended with some variability from the soma,                 length (Fig. 8g), although the apical dendrites were clearly
either favoring radial processes that descended toward the              truncated by sectioning.
white matter or more lateral, tufted projections. Dendrites                Greater variation in apical dendritic orientation was
possessed a relatively low density of spines, with a DSD                observed with horizontal and inverted pyramidal neurons.
ranging from 0.30 to 0.36 (Table 2). Sholl analysis                     Horizontal pyramidal neurons (n = 2, Fig. 6I–J), found in
revealed two apical dendritic peaks, with the second peak               layer III of the occipital lobe, possessed a narrow, elon-
being somewhat larger (Fig. 8e). Basilar dendritic density              gated soma from which a quickly ramifying apical dendrite
peaked closer to the soma than did the apical dendrite.                 extended laterally or obliquely (\45"). Basilar dendrites

                                                                                                                              123
Brain Struct Funct

Fig. 6 Neurolucida tracings of spiny neurons in the elephant              neurons (F–H); horizontal pyramidal neurons (I–J); multiapical pyrami-
occipital cortex presented to indicate their relative soma depths         dal neurons (K–L); flattened pyramidal neurons (M–N, P). The quanti-
from the pial surface (in lm): inverted pyramidal neuron (A); crab-like   tative dependent measures for all neurons are provided in Table 2. Scale
neurons (B, O); superficial pyramidal neurons (C–E); magnopyramidal       bar 100 lm

branched into distinct, parallel collaterals or extended                  neurons (n = 3, Fig. 6M–N, P) were found in deep layer
numerous secondary and tertiary dendrites. DSD was                        III and upper layer VI of the occipital cortex, where they
between 0.44 and 0.47 for the two cells (Table 2). In the                 appeared to be fairly common. Flattened pyramidal neu-
Sholl analysis, the apical and basilar dendrites were                     rons sent multiple, thick apical dendrites from opposing
roughly equal in density and length (Fig. 8f). Inverted                   ends of the soma to ascend outward and then gradually
pyramidal neurons (n = 3, Figs. 4A, C, 5G–H, 6A), found                   toward the pial surface. Somata varied from elongated to
in layer III of frontal and occipital cortices, possessed                 globular. Dendrites extended further laterally than verti-
apical dendrites that descended and bifurcated from a                     cally, with basilar dendrites often descending toward the
roughly triangular soma. In the frontal lobe, the length of               white matter alongside apical dendrite collaterals. Their
the apical process varied from 600 to 1,100 lm and pos-                   thick dendrites made the cells, by volume, among the
sessed few collaterals. The basilar skirt projected toward                largest in the present sample. Their DSD was similar to
the pial surface, producing a circular receptive area. The                other large pyramidal cells, ranging from 0.37 to 0.41
smallest and most superficial of these cells (Fig. 6A) pos-               (Table 2). Sholl analysis revealed the maximum density of
sessed a short apical shaft that descended obliquely for less             apical dendrites to be further from the soma than that of the
than 400 lm in length, and an axon that emerged from the                  basilar dendrites (Fig. 8d). Crab-like neurons (n = 3,
base of the soma and bifurcated, one branch ascending and                 Figs. 5A, 6B, O) were so designated because their dendritic
the other descending. DSD ranged from 0.39 to 0.46                        branches emerged from opposite ends of a rounded soma to
(Table 2). Sholl analyses confirmed the inverted pyramidal                ramify symmetrically, producing claw-like shapes. These
neuron’s (Fig. 8b) structural similarity to the magnopyra-                were located in layers II and III and at the junction between
midal neurons.                                                            layer V and VI of the frontal and occipital lobes. DSD
   Finally, two ‘‘atypical’’ spiny neuron variants appeared               varied greatly, ranging from 0.39 to 0.67 (Table 2). Sholl
to have horizontally oriented dendritic arbors: flattened                 analysis indicated that crab-like neurons (Fig. 8j) exhibited
pyramidal and ‘‘crab-like’’ neurons. Flattened pyramidal                  a relatively small dendritic receptive area.

123
Brain Struct Funct

Fig. 7 Neurolucida tracings of aspiny neurons found in elephant          (A–D, J–K); bipolar neurons (E–I); neurogliaform neuron (L). The
frontal (A–I) and occipital (J–L) cortices presented to indicate their   quantitative dependent measures for all neurons are provided in
relative soma depths from the pial surface (in lm): multipolar neurons   Table 2. Scale bar 100 lm

Aspiny neurons                                                              The neurogliaform neuron (n = 1, Fig. 7L), also called a
                                                                         spiderweb cell by Ramón y Cajal (1922), was an aspiny,
Aspiny multipolar neurons (n = 6, Figs. 7A–D, J–K) were                  diffusely branching multipolar cell in layers II and III of
found throughout all layers of frontal and occipital cortices.           occipital cortex. Three main dendritic branches emerged from
They varied in appearance, but generally exhibited a small,              an elongated soma and gave rise to numerous, approximately
globular soma from which a few (2–6) primary, straight den-              200 lm dendritic projections that bifurcated densely in all
dritic projections extended to 1,000 lm in length. In the Sholl          directions to produce a spherical, net-like receptive area with a
analysis, some multipolar cells (Fig. 8h) exhibited dendritic            TDL that far surpassed that of any other cell (Table 2). A few
projections that were longer than any other cells, excluding the         longer dendrites (up to 600 lm in length) extended along the
magnopyramidal and inverted pyramidal neurons.                           cell’s vertical axis. This cell had by far the densest dendritic
   Bipolar neurons (n = 5, Fig. 7E–I), similar in size and               arbor (DSC = 337), with over twice as many ring intersec-
shape to the multipolar neurons, were found in layers III                tions as any other neuron in the Sholl analysis (Fig. 8k).
and V of frontal cortex. They possessed a small, globular                   Quantitatively, three general observations about aspiny
soma from which two main, thick dendritic segments                       interneurons and spiny neurons in the elephant emerge
extended in opposite directions. Thin collaterals, up to                 (excluding the neurogliaform neuron, which was a clear
900 lm in length, extended outward from these segments                   dendritic outlier): (1) aspiny neurons in frontal cortex
and ramified to produce a roughly symmetrical bitufted                   (MVol = 11,870 ± 5,337 lm3; MTDL = 4,650 ± 1,462 lm)
appearance. According to the Sholl analysis, the dendrites               appeared somewhat larger than those in occipital cortex
of these cells (Fig. 8i) were, on average, shorter and less              (MVol = 7,514 ± 2,441 lm3; MTDL = 2,558 ± 599 lm),
dense than the comparable multipolar neurons.                            although the small sample size severely limits the value of

                                                                                                                               123
Brain Struct Funct

Table 2 Summary statistics for elephant neurons in frontal and occipital cortex
Type                               Cella         Vol.b          TDLc         MSLc   DSCd   DSNe    DSDf   Soma sizeg      Soma depthh

Frontal
    Crab-like                      5A             5,732          4,249        83     51    2,863   0.67    439             424
    Superficial pyramidal          5B             8,233          4,574        82     56    2,036   0.45    286             741
    Superficial pyramidal          5C            17,880          6,740        95     71    4,162   0.62    478             691
    Superficial pyramidal          5D            30,300          7,603       104     98    4,839   0.64    479             818
    Superficial pyramidal          5E            17,250          8,233       106     78    5,379   0.65    506             875
    Fork                           5F            55,256          8,050       117     69    4,685   0.58   1,154           1,436
    Inverted pyramidal             5G            13,937          5,634        95     59    2,608   0.46    681            1,101
    Inverted pyramidal             5H            18,055          7,370        93     79    3,012   0.41    706            1,203
    Matriarch                      5I            27,149          7,716       100     77    2,489   0.32   1,061           1,066
    Matriarch                      5J            33,500         10,838       111     97    5,856   0.54    766            1,299
    Matriarch                      5K            43,904          8,269       115     72    3,977   0.48    973            1,317
    Matriarch                      5L            28,360         10,377       113     92    5,430   0.52    799            1,423
    Multiapical pyramidal          5M            54,261          6,742        98     69    2,340   0.35   1,552           1,662
    Multipolar                     7A            10,843          6,879       186     37       –    –       387            1,124
    Multipolar                     7B             9,311          5,152       152     34       –    –       611            1,069
    Multipolar                     7C            11,905          3,778       114     33       –    –       801            1,811
    Multipolar                     7D            24,569          5,839       195     30       –    –      1,375           2,027
    Bipolar                        7E            13,497          5,883       115     51       –    –       757            1,811
    Bipolar                        7F             9,186          4,129       118     35     193    0.05    494            1,132
    Bipolar                        7G             8,510          3,996       125     32       –    –       351            1,203
    Bipolar                        7H             5,851          1,946        97     20       –    –       411            1,893
    Bipolar                        7I            13,160          4,248       112     38       –    –       737            1,641
Occipital
    Inverted pyramidal             6A             5,173          3,465        96     36    1,346   0.39    398             823
    Crab-like                      6B             4,596          2,500       104     24    1,214   0.49    407             806
    Superficial pyramidal          6C            10,897          5,921        82     72    3,346   0.57    505             722
    Superficial pyramidal          6D            16,264          5,509       115     48    3,630   0.66    643             746
    Superficial pyramidal          6E            11,312          6,752        83     81    3,914   0.58    256             650
    Magnopyramidal                 6F            18,924          8,752       115     76    4,094   0.47    780            1,470
    Magnopyramidal                 6G            76,591          7,925       113     70    3,148   0.40   2,015           1,487
    Magnopyramidal                 6H            46,090          6,895        90     77    2,296   0.33   2,004           1,525
    Horizontal pyramidal           6I            29,131          7,554       113     67    3,521   0.47    988            1,571
    Horizontal pyramidal           6J            28,811          6,411        88     73    2,853   0.44    941            1,090
    Multiapical pyramidal          6K            35,055          4,428        84     53    1,311   0.30   1,562           1,536
    Multiapical pyramidal          6L            32,114          6,375       133     48    2,291   0.36   1,284           1,734
    Flattened pyramidal            6M            58,109          6,187       115     54    2,293   0.37   1,694           1,456
    Flattened pyramidal            6N            36,137          4,346        92     47    1,706   0.39   1,222           1,407
    Crab-like                      6O            13,679          4,938        60     83    1,949   0.39    689            2,003
    Flattened pyramidal            6P            57,494          6,476        93     70    2,674   0.41   1,092           2,233
    Multipolar                     7J             9,241          2,981       110     27     210    0.07    479             672
    Multipolar                     7K             5,788          2,134        71     30      34    0.02    423             704
    Neurogliaform                  7L            26,959         14,818        44    337       –    –       857             528
a
    Refers to tracings of individual cells as indentified in Figs. 5, 6, 7
b
    Volume in lm3
c
    Length in lm
d
    Number of segments per neuron
e
    Number of spines per neuron
f
    Number of spines per lm of dendritic length
g
    Soma size in lm2
h
    Soma depth in lm from the pial surface

123
Brain Struct Funct

Fig. 8 Sholl analyses of 11 cell types assessing the relative            The apical dendrite either tended to extend for a much larger distance
complexity of basilar, apical, and total dendritic branching patterns.   than the basilar dendrites (a, b) or was nearly equal to them (c–g).
Concentric rings, separated by 20 lm increments and centered on the      Aspiny neurons (h–j) generally had low dendritic densities, the
soma, were used to measure dendritic intersections. Neurons a–g were     notable exception being the neurogliaform neuron (k), which had the
spiny. Neurons h–k were aspiny. a–c exhibited relatively long apical     highest dendritic density of all neurons. Asterisks represents that the
dendrites, d–g had shorter apical dendrites. For cells with basilar      y-axis for the neurogliaform graph (k) is set to a larger scale than
dendrites, the density of basilar intersections peaked before 200 lm,    other cells to compensate for its particularly high dendritic density
as was the case for the total dendritic density of all aspiny neurons.

this comparison; (2) the aspiny interneurons (MVol =                     soma size and volume (r(40) = 0.353, p \ 0.05), which
11,078 ± 5,147 lm3; MTDL = 4,270 ± 1,569 lm) exhib-                      underscored the importance of maintaining similar soma
ited substantially less overall dendritic extent than the                depths between areas for subsequent comparisons.
spiny neurons (MVol = 28,765 ± 11,078 lm3; MTDL =                           In terms of appearance, superficial pyramidal neurons
6,580 ± 1,906 lm); and (3) the aspiny interneurons                       exhibited typical basilar skirts, which projected radially
exhibited considerably longer (MMSL = 127 ± 37 lm),                      for a short distance from the soma. They were charac-
but fewer (MDSC = 33 ± 8) dendritic branches than did the                terized by apical dendrites that generally bifurcated
spiny neurons (MMSL = 100 ± 15 lm; MDSC = 67 ± 17).                      around 50 lm from the soma to form two, ascending
                                                                         apical shafts (Figs. 5D–E, 6C, E, 10a, d), or by two
Superficial pyramidal neurons                                            diverging apical dendrites (Figs. 5C, 6D) that emerged
                                                                         directly from the soma. These resulting apical trunks
Superficial pyramidal neurons (n = 40; Figs. 4B, E, 5B–E,                traveled obliquely to the pial surface, thus creating a large
6C–E) were well impregnated across both occipital and                    apical arbor for each cell and a crisscross or ‘‘V’’
frontal cortices (Fig. 10), facilitating a more detailed                 appearance in the upper layers (Fig. 11b–c). Generally,
quantitative analysis. These neurons were sampled pri-                   these branches appeared to bundle with other branches
marily from upper cortical layer III with minimal variation              traveling at similar angles toward the pial surface
of soma depth across cortical areas (frontal 818 ± 108 lm;               (Fig. 11a, d).
occipital 769 ± 149 lm). Measurements of soma size
indicated minimal variation between cortical areas (frontal              Dendritic morphology-dependent measures
515 ± 80 lm2; occipital 517 ± 156 lm2). There was a
significant positive correlation between soma depth and                  Selected dependent measures for basilar data were first
soma size (r(40) = 0.451, p \ 0.01), as well as between                  analyzed using a MANOVA test statistic. Due to direct

                                                                                                                                    123
Brain Struct Funct

Fig. 9 Photomicrographs of two magnopyramidal-taproot or matri-            tracings can be seen in Fig. 5L for a, and in Fig. 5I for b. Scale bar
arch neurons (a, b) with higher magnification insets of their spine-rich   100 lm for a and b, and 25 lm for the insets
ascending apical (a) and descending taproot (t) dendrites. Neurolucida

associations between DSD and DSC with other measures,                      was in the opposite direction for MSL, but was not sig-
these variables were not used in the MANOVA. The                           nificant (Fig. 13c). Both TDL (by 19.0%) and MSL (by
MANOVA was significant (F(4, 35) = 3.125, p \ 0.05,                        5.0%) were greater for apical dendrites in the frontal cortex
g2p = 0.263). Post hoc analyses, consisting of six one-way                 than in the occipital cortex. Descriptive data from complete
ANOVAs, were then conducted for each dependent vari-                       neurons (i.e., basilar and apical dendrites combined)
able. All dependent measures, except MSL and DSD, were                     showed that, overall, TDL was greater (by 13.5%) and
significantly higher in frontal cortex, indicating the pres-               MSL was shorter (by 9.2%) in the frontal cortex than in the
ence of more complex basilar dendrites. Although there                     occipital cortex.
were not enough complete apical branches to warrant sta-
tistical tests, descriptive data indicated a similar pattern for           Dendritic segment count
apical dendrites. The differences in complexity can be seen
visually in Fig. 12, which contains representative tracings                There was a significant difference in basilar DSC across
from each region.                                                          both areas (F(1, 38) = 10.93, p \ 0.05, g2p = 0.223), with
                                                                           basilar dendrites in frontal cortex exhibiting 25.1% higher
Volume                                                                     DSC than occipital cortex (Fig. 13d). Apical dendrites also
                                                                           appeared to have higher DSC in frontal cortex (by 14.8%),
Basilar dendritic volume was significantly greater (F(1, 38)               and complete neurons in frontal cortex had 20.7% higher
= 5.80, p \ 0.05, g2p = 0.132) in frontal than in occipital                DSC than those in occipital cortex.
cortex (by 25.2%; Fig. 13a). On average, neurons from
frontal cortex contained 29.5% thicker apical dendrites and                Dendritic spine number and density
23.3% more overall volume than neurons from occipital
cortex.                                                                    Basilar dendrites of neurons in the frontal cortex displayed
                                                                           a significantly higher DSN (by 25.1%; Fig. 13e) than
Dendritic length                                                           neurons in the occipital cortex (F(1, 38) = 7.95, p \ 0.01,
                                                                           g2p = 0.173). DSD was not significant (Fig. 13f). Apical
Basilar dendrites in frontal cortex exhibited significantly                dendrites in frontal cortex contained 29.4% more spines
greater TDL (F(1, 38) = 7.15, p \ 0.05, g2p = 0.158) than                  and had a 13.4% higher DSD than those in occipital cortex.
basilar dendrites in occipital cortex (Fig. 13b). The trend                Evidence from complete neurons indicated that frontal

123
Brain Struct Funct

Fig. 10 Photomicrographs of Golgi-impregnated superficial pyrami-   magnifications. Note in both cortical regions the presence of
dal neurons in occipital (a) and frontal (d) cortices. Associated   bifurcating apical dendrites. For a, b, d, f scale bars 50 lm. For c,
photomicrographs of apical (occipital: b; frontal f) and basilar    e scale bars 25 lm
(occipital: c; frontal e) dendrites are also shown at higher

                                                                                                                             123
Brain Struct Funct

Fig. 11 Photomicrographs of V-apical bundling within elephant           to bundle together as they obliquely approach the pial surface. For
occipital cortex. At lower magnifications (b and c), a V-shaped         a and b scale bars 50 lm. For c scale bar 100 lm. For d scale bar
crossing pattern resulting from bifurcating apical dendrites is         25 lm
apparent. At higher magnifications (a and d), apical dendrites appear

superficial neurons contained 20.2% more spines and had a               (frontal 4.88 ± 0.50 intersections; occipital 4.10 ± 0.48
7.6% higher DSD than did occipital neurons.                             intersections).

Sholl analysis                                                          Comparison to human supragranular pyramidal neurons

Frontal and occipital regions had similar shapes in Sholl               Elephant superficial pyramidal neurons were compared to
analysis profiles, but frontal results were slightly                    supragranular pyramidal neuron data gathered with similar
higher on average (Fig. 14a, b). Basilar dendrites in both              methodology in human inferior frontal (specifically, area
areas reached their maximum complexity (frontal                         11, from Jacobs et al. 2001) and occipital regions (area 18,
16.45 ± 0.99 intersections; occipital 12.80 ± 0.86 inter-               from Jacobs et al. 1997). These regions were chosen to be a
sections) 120 lm from the soma, and apical dendritic                    relatively close topographic match for those examined in
complexity peaked around 200 lm from the soma                           the elephant. The human supragranular pyramidal neurons

123
Brain Struct Funct

Fig. 12 Sample tracings of superficial pyramidal neurons, and their       cortical area, the two neurons with relatively complete apical
corresponding dendritic measurements, from frontal (cells a–d) and        dendrites appear on the left. Note also the bifurcation of all apical
occipital (cells e–h) cortices. Neurons are presented to indicate their   dendrites. In general, frontal neurons appeared more dendritically
relative soma depths from the pial surface (in lm). Within each           complex than occipital neurons. Scale bars 100 lm

(n = 116) were traced in 6 neurologically normal subjects                 elephant occipital cortex, basilar dendrites exhibited 52.3%
(4 men, 23–69 years of age; 2 women, 32 and 34 years of                   higher DSN and 49.9% higher DSD than in humans.
age).
                                                                          Sholl analysis
Dendritic length and segment count
                                                                          Although Sholl analysis profiles were similar across both
On average, elephant superficial pyramidal neurons dem-                   area 11 and area 18 in human, they varied considerably
onstrated only slightly higher TDL than human neurons                     from those in the elephants (Fig. 14). On average, elephant
both in frontal (by 6.8%) and in occipital (by 2.9%) cor-                 basilar dendrites extended 760 lm from the soma in frontal
tices (Fig. 15a). However, TDL was manifested differently                 cortex and 740 lm from the soma in occipital cortex. In
in humans and elephants, as revealed by MSL and DSC                       contrast, human basilar dendrites extended 380 lm from
values. Basilar dendrites in the elephant exhibited greater               the soma in the frontal cortex and 400 lm in occipital
MSL both in frontal (by 23.3%) and in occipital cortices                  cortex. At the high point of dendritic intersections, human
(by 35.6%; Fig. 15b) than they did in humans. In contrast,                supragranular pyramidal neurons contained more intersec-
elephant DSC was considerably lower than human DSC                        tions than did elephant superficial pyramidal neurons. In
(by 16.0% in frontal cortex, and by 33.5% in occipital                    frontal cortex, human basilar dendrites peaked at 22.90 ±
cortex; Fig. 15c).                                                        0.90 intersections, whereas elephant dendrites peaked at
                                                                          16.45 ± 0.99 intersections. In occipital cortex, human
Dendritic spines                                                          basilar dendrites peaked at 21.47 ± 0.65 intersections,
                                                                          whereas elephant dendrites peaked at 12.80 ± 0.86 inter-
Elephant superficial pyramidal neurons exhibited greater                  sections. Overall, the Sholl analysis profiles confirm that
values for DSN and DSD than did human neurons. In                         human supragranular pyramidal neurons were more con-
elephant frontal cortex, basilar dendrites exhibited 63.6%                densed and contained more branch points than the elephant
higher DSN and 61.7% higher DSD than in humans. In                        superficial pyramidal neurons.

                                                                                                                                   123
Brain Struct Funct

Fig. 13 Bar graphs of relative volume (a), total dendritic length (b),   basilar dendrites in frontal cortex had significantly greater volume,
mean segment length (c), dendritic segment count (d), dendritic spine    TDL, DSC, DSC, and DSN than did those in occipital cortex. Error
number (e), and dendritic spine density (f) of superficial pyramidal     bars represent SEM
neuron basilar dendrites in elephant frontal and occipital cortex. The

Discussion                                                               and contains a variety of pyramidal and non-pyramidal
                                                                         neurons as seen in many other mammalian species. This
The present study is the first to document neuronal                      cortical lamination pattern with lack of an internal gran-
morphology in the African elephant cortex. In terms of                   ular layer IV may reflect a particular organization of
cytoarchitecture, the elephant neocortex is characterized                cortical connectivity in elephants, perhaps comparable to
by the presence of only five cortical layers as it altogether            that in cetaceans, where agranularity is also a neocortical
lacks a visible layer IV, which is quite different from                  characteristic (Hof et al. 2005; Hof and Van der Gucht
what is observed in primates. Although it has not been                   2007). Although, we are currently exploring the cytoar-
studied in much detail to date, especially in terms of                   chitecture of elephant cortex in detail, it remains clear
regional differences, we can make several generalizations.               that complex neocortical gyrification and distinct laminar
Layer I appears to be relatively thick and more cellular                 organization represent remarkable features of the elephant
than in most terrestrial species. Layer II is clearly visible,           brain.
densely packed, and displays cellular clustering in many                    In terms of neuromorphology, elephant cortex featured a
cortical regions, particularly in the insula (Hakeem et al.              great diversity of large, complex neurons, with consider-
2009), as has been reported in cetaceans (Manger et al.                  able variety exhibited by ‘‘atypical’’ spiny neurons. A
1998; Hof and Van der Gucht 2007). Layer II also con-                    prominent characteristic in elephant cortex was the
tains occasional large pyramid-like neurons. Layer III is                V-shaped arrangement of bifurcating apical dendrites.
relatively thick and populated by large pyramidal neurons,               Quantitatively, the dendrites of superficial pyramidal neu-
the density and size of which clearly vary among different               rons in the elephant frontal cortex were more complex than
cortical domains; this fact deserves a detailed investiga-               those in occipital cortex. In comparison to humans, ele-
tion. Layer V presents as a relatively thin row of very                  phant superficial pyramidal neurons exhibited similar
large pyramidal cells usually distributed in small groups,               overall basilar dendritic length, but individual dendrites
presumably representing principal efferent neurons, and a                tended to be longer in the elephant with less intricate
thicker deep portion containing smaller pyramidal cells.                 branching. Before considering the implications of these
Layer VI is of variable thickness among cortical regions                 findings, methodological issues must be addressed.

123
Brain Struct Funct

Fig. 14 Graphic representation of Sholl analysis results for elephant   (d) cortices. Although dendritic extent between areas and species is
superficial and human supragranular pyramidal neurons, showing          similar, elephant dendrites are distributed across a greater distance
mean numbers of dendritic intersections in elephant frontal (a),        from the soma and exhibit fewer intersections at the peak than
elephant occipital (b), human frontal (c), and human occipital          observed in humans. Error bars represent SEM

Methodological considerations                                           premotor activities. The function of the occipital cortex is
                                                                        even more problematic, as this region could support visual
In general, the same constraints that apply to Golgi-stained            or even auditory processes. As such, it is impossible to
human tissue also apply here (Jacobs and Scheibel 2002):                correlate dendritic/spine measure with the functional
(1) a small sample size in terms of subjects and sampled                attributes of these regions, as has been done in primates
neurons (Jacobs and Scheibel 1993); (2) lack of historical              (Elston and Rosa 1998a, b; Jacobs et al. 2001). Moreover,
information on the subjects (Jacobs et al. 1993); (3)                   it is not possible to make definitive statements about the
underestimation of spines in light microscopy (Horner and               regional distribution of traced neurons in the present study
Arbuthnott 1991); and (4) inherent issues with Golgi                    despite observing that some neuron types appeared only in
impregnations (Braak and Braak 1985). These general                     one cortical region.
limitations are accepted and we focus here on three broader
issues.                                                                 Neuromorphological nomenclature

Functional classification of elephant cortex and regional               Unambiguously classifying neurons is problematic when
specialization                                                          the criteria are not clearly defined (Bota and Swanson
                                                                        2007), are too restrictive (Germroth et al. 1989), or when
Functional characteristics of the sampled regions could not             cellular morphologies form a continuum rather than distinct
be determined. The frontal cortex, although located in what             categories, as may be the case for extraverted telencephalic
appear to be anterior orbital gyri, could subserve prefrontal           neurons (Sanides and Sanides 1972; Ferrer and Perera
executive functions (Barbas 1995) or, equally likely,                   1988). Further classification issues emerge when definitions

                                                                                                                                 123
Brain Struct Funct

Fig. 15 Bar graphs displaying the differences in relative total        elephant and human pyramidal neurons show similar TDL values,
dendritic length (a), mean segment length (b), and dendritic segment   human neurons contain shorter segments and more intricate branch-
count (c) for basilar dendrites in elephant superficial pyramidal      ing. All human data were obtained from Jacobs et al. (1997) for
neurons and human supragranular pyramidal neurons across frontal       occipital cortex (specifically area 18), and Jacobs et al. (2001) for
and occipital cortex. Sample tracings of elephant superficial and      frontal cortex (specifically area 11). Error bars represent SEM
human supragranular pyramidal neurons are also provided. Although

change over time, as typified by the term ‘‘pyramidal,’’               limits our overall conclusions regarding neuronal types.
which now applies to a much broader category of neurons                Finally, the present study was also limited because it
than it did originally (Masland 2004). Compounding the                 described neurons based only on somatodendritic measures.
classification problem is an essentially Euarchontoglires-
centric nomenclature in the literature, whereby the                    Comparison of human and elephant dendritic/spine
(anthropoid) primate together with the (murid) rodent brain            measures
constitutes the template against which all other brains are
compared, thus dictating what is ‘‘typical’’ versus ‘‘atypi-           With regard to dendritic measures, only portions captured
cal’’ (Manger et al. 2008). In the present study, we have              in the 120-lm thick section could be compared. Because
attempted to follow existing nomenclature where possible               the basilar dendrites in elephants exhibited greater TDL
or, when appropriate, named cells based loosely on their               than in humans, the observed TDL difference between the
dendritic morphology (e.g., crab-like neuron). Although we             two species was probably attenuated (Jacobs et al. 1997).
observed many examples of the presented neurons in our                 Nevertheless, sampling within an equal volumetric ‘‘slice’’
examination of the Golgi-stained sections, we only traced              provides useful relative measures for comparison. With
those that were relatively complete and unobscured, which              regard to spine measures, several issues prohibit a

123
Brain Struct Funct

meaningful interpretation of the relative DSN and DSD          subsequent investigations in several mammals (e.g., pri-
values. First, the human brains were immersion-fixed,          mates, carnivores, ungulates, and rodents) have concluded
whereas the elephant brains were perfusion-fixed, the latter   that it is a type of enveloping neuron (or Umfassungszelle)
revealing greater detail (i.e., spines; Morest and Morest      and thus is merely a variant of pyramidal neurons (Juba
2005). Second, autolysis time in the humans was longer         1934; Syring 1956). Finally, elephant magnopyramidal,
(*14 h) than in the elephants (*2 h), which potentially        multiapical, and fork neurons appear at least as dendritic-
decreased the number of impregnated spines in the human        ally complex as infragranular pyramidal neurons in the cat
samples (de Ruiter 1983). Finally, although data were          parietal cortex (Yamamoto et al. 1987), human cingulate,
collected in a methodologically similar manner, human          and motor cortices (Meyer 1987; Schlaug et al. 1993).
neurons were traced using a Neurolucida Lucivid system,           Within a more polymorphic subgroup of pyramidal
whereas elephant neurons were traced on a Neurolucida          neurons are the horizontal and inverted pyramidal neurons
camera system. It was recently discovered that the higher      (Ramón y Cajal 1891; Van der Loos 1965). In terms of
magnification in the camera system increases the number        morphology, these are similar to magnopyramidal and
of spines identified by about 17% (Anderson et al. 2010),      multiapical neurons, but are slightly smaller in dendritic
clearly making accurate spine comparisons between the          extent and differ in their specific orientations. Although
two species problematic.                                       horizontal pyramidal neurons have typically been noted in
                                                               superficial and deep cortical layers in primates and rodents
Spiny neurons                                                  (Meyer 1987; Miller 1988), elephant horizontal pyramidal
                                                               neurons were located in layer III. These strongly resembled
The elephant cortex is characterized by an even greater        one variant of the asymmetrical pyramidal neurons in
variety of spiny neurons than the morphologically hetero-      inferior temporal cortex (Fig. 8e of De Lima et al. 1990),
geneous inferior temporal cortex of macaque monkeys and        although one could argue that what De Lima et al. refer to
rats (Germroth et al. 1989; De Lima et al. 1990). Although     as a prominent, laterally oriented, basilar dendrite is actu-
several of these neuron types have been observed in other      ally the apical dendrite. The elephant variant appears to be
eutherian species, what remains noteworthy in the elephant     a more dendritically complex version than what has been
is their arrangement, which represents a striking departure    noted in layer VI of both lissencephalic and gyrencephalic
from the vertical apical dendrites and cortical columns that   brains (Ferrer et al. 1986a, b, 1987). Although the function
have been deemed the fundamental, if not canonical             of these neurons remains unclear, it is possible that they
building block of the cerebral cortex (Mountcastle 1997;       contribute to the lateral integration of synaptic input (Van
Innocenti and Vercelli 2010).                                  Brederode et al. 2000), which has been suggested as a
    These spiny cortical neurons exist along a continuum       major step in neocortical evolution (Ferrer et al. 1986b).
from those that appear more pyramid-like to those that         Much the same as horizontal pyramidal neurons, elephant
radically differ from the ‘‘typical’’ pyramidal neuron in      inverted pyramidal neurons were located in layer III. In
terms of morphology and/or orientation. In the elephant,       human temporal cortex, they have been documented in
three neuron types, in addition to superficial pyramidal       layer V (Ong and Garey 1990). In chimpanzees, inverted
neurons, approximate general pyramidal neuron morphol-         pyramidal neurons are located in layers III, V and espe-
ogy: magnopyramidal, multiapical, and fork neurons. The        cially VI of sensorimotor cortices, constituting a small
only exact comparison point in the literature is the single    percentage (\1%) of all pyramidal neurons (Qi et al. 1999).
Indian elephant neuron depicted in Barasa and Shochato-        In rats, rabbits, cats, sheep, sloths, anteaters, rock hyrax,
vitz (1961), which is consistent with the present layer V      and elephant shrews, they also tend to be located in in-
magnopyramidal neurons. Although pyramidal neurons in          fragranular layers and constitute a small (1–8.5%) per-
the cat, dog, pig, and sheep resemble those in the elephant,   centage of all neurons in the cortex (Parnavelas et al. 1977;
the most striking similarities are found in the cow, horse,    Mendizabal-Zubiaga et al. 2007; Sherwood et al. 2009).
two-toed sloth, and anteaters (Barasa 1960; Ferrer et al.      These excitatory neurons have been shown to project to
1986b; Sherwood et al. 2009), which exhibit prominent,         ipsi- and contra-lateral cortical regions, as well as to the
bifurcating apical dendrites much the same as those            claustrum and striatum (Bueno-López et al. 1991; Mendi-
observed in elephant magnopyramidal and multiapical            zabal-Zubiaga et al. 2007). In the elephant, however, the
neurons. Unfortunately, size comparisons with cow and          extent to which inverted pyramidal neurons share connec-
horse neurons are not possible because Barasa (1960) did       tional and/or functional attributes with their counterparts in
not provide scale bars. The fork neuron has been observed      other species remains unclear.
in the insula and hippocampus of humans, chimpanzees,             One type of elephant neuron that does not have a clear
and orangutans (Ngowyang 1936; de Crinis 1934).                counterpart in the literature is the magnopyramidal-taproot
Although originally considered a unique type of neuron,        or ‘‘matriarch’’ neuron of the frontal cortex. Although they

                                                                                                                  123
You can also read